Interview
summary
Argues that real progress comes from concrete model abilities not broad "AI" labels and demonstrates where deep learning wins on visual feature extraction but struggles on multi-step reasoning. Focuses on using memory and attention mechanisms to scale models to harder tasks and pushes for continual learning without per-task retraining so a single model improves over time. Notes phones already run many ML systems and contrasts marketing hype with technical limits.
AI-written summary · about
chapters (4)
related talks
04mFrom big data to human-level artificial intelligence
46mAI and ML, separating hype from what actually matters
36mKeynote: AI, ML, Cloud for ML
03mRethinking software engineering in the AI era
26mKeynote: AI in the Real World: Today and Tomorrow
27mThe Cutting Edge of AI: From LLMs to Multimodal Intelligence
13mArtificial General Intelligence: Why aren't we there yet?
42mReasonable AI